[AfA2026-PhysicalAI] AI-Enabled Helmet Detection–Based Safety Interlock System for Two-Wheelers

Power Bank

Arduino® UNO™ Q 2GB

Webcam, Logitech® HD Pro

Modulino® Latch Relay

Modulino® Buzzer

USB Hub, Bus Powered

Arduino App Lab

<html><body>Road accidents involving two-wheelers account for a significant portion of traffic fatalities, especially in developing countries. A major cause of severe injury and death is non-usage of helmets, despite legal mandates. Existing solutions rely on enforcement and awareness, which are often insufficient. There is a need for an intelligent, automatic system that ensures helmet compliance before the vehicle can be started, without relying on human intervention. This project presents an AI-based safety interlock system for two-wheelers that prevents the vehicle from starting unless the rider is wearing a helmet. A camera mounted facing the rider captures real-time images when the ignition key is inserted. A computer vision model detects whether a helmet is being worn. If a helmet is detected → Relay closes, allowing ignition, If not → Buzzer remains ON and relay stays open, preventing the bike from starting. The system is low-cost, real-time, and user-friendly, making it suitable for mass deployment. AI Model Used: This project uses a pre-trained model to helmet on a live video feed from a camera. The workflow involves continuously getting the frames from a USB camera, processing it through an AI model using the video_imageclassification Brick, and displaying the classification along with their corresponding probabilities. Software Architecture: The software architecture follows a modular and event-driven design, enabling efficient communication between components and easy scalability. Application Logic Flow: <ol><li>System initializes when the ignition key is inserted </li><li>Camera continuously captures rider images </li><li>AI inference module processes image frames </li><li>Detection results are passed to the control logic </li></ol>Control logic drives: <ol><li>Relay (ignition enable/disable) </li><li>Buzzer (user feedback) </li></ol>Component Communication: <ol><li>Camera → AI Inference Module: Image frames </li><li>AI Model → Decision Logic: Helmet detection result </li><li>Decision Logic → Actuators: GPIO control signals for relay and buzzer </li></ol>This architecture allows: Real-time response; Easy model upgrades; Seamless integration of additional safety features in the future. Social Impact: <ol><li>Reduces head-injury-related fatalities </li><li>Encourages helmet compliance without enforcement </li><li>Suitable for developing countries </li><li>Supports SDG 3: Good Health and Well-Being </li></ol>Unlike RFID-based or helmet-wired systems, this solution: does not require any modifications to the helmet, works with any standard helmet, and uses vision-based AI, making it more flexible and scalable while continuously ensuring that the rider is wearing a helmet for the entire duration that the bike remains in operation. <figure><img src="https://projects.arduinocontent.cc/0735a23f-e2f9-4752-8504-3349ddba909d.png"/><figcaption>Schematic Diagram</figcaption></figure>User Interface &amp; Feedback: The system interacts with the user through simple and intuitive physical feedback mechanisms, eliminating the need for displays or mobile applications. Buzzer (Auditory Feedback) : A buzzer provides immediate audio alerts when the rider is not wearing a helmet. The buzzer remains active until a helmet is detected, prompting the rider to comply. Ignition Interlock (Relay Feedback) : The relay acts as a physical safety lock on the ignition circuit. When a helmet is detected, the relay closes and allows the vehicle to start. If no helmet is detected, the relay remains open, preventing ignition. This minimal interface ensures: <ol><li>Zero distraction for the rider </li><li>Clear and instant feedback </li><li>High reliability in real-world riding conditions </li></ol><figure><iframe src="https://youtu.be/5eot_byg3oc"></iframe></figure></body></html>

Github Repository

sketch.ino

main.py

[AfA2026-PhysicalAI] AI-Enabled Helmet Detection–Based Safety Interlock System for Two-Wheelers

No helmet, no ignition — powered by real-time edge AI.